Dynamic features clustering for object detection and tracking in interactive perception

In order to be able to interact with its environment and solve non-trivial object-based tasks (e.g. manipulation), a robot must be able to locate objects in its perceptual field, and to track them throughout the interaction. In the case of a static task and structured environment, for example objects on a tabletop, those perceptual abilities can be hardcoded. But in an open-ended context where the environment, the task and the objects can't be known in advance and can change during the interaction, it is desirable for the robot to be able to bootstrap its perceptual abilities with limited assumptions, and to be able to update its interpretation of the visual scene during the interaction to keep track with the changes in the setup (e.g. new objects, changes in the background, ...).

The goal of this internship is to build an interactive perception system to recognize moveable object through interactions of a robot with its environment. It is based on the PhD of Leni Le Goff and on the master thesis of Elias Hanna. The proposed approach will be tested on some of ISIR robotic platforms as the Baxter robot and/or the PR2 robot.

More details on the goals and on the candidate profile in the PDF description.

Alexandre Coninx
Stéphane Doncieux
Référent Universitaire: 
Fichier Descriptif: 
2 020

User login